2020 年 5 巻 1 号 p. 4-21
In recent years, there have been attempts to generate stories automatically by utilizing computers and various other kinds of information technology. However, it remains unclear how story sequences can be made to look natural and appear interesting. This study conducted a data description of 134 famous Japanese comic detective stories and listed 37 types of plot elements, 10 types of tricks, 9 types of criminal motives, and 10 types of relationships between victims and criminals, in order to realize a computational narratological analysis. Based on these parameters, eleven factors have been extracted through factor analysis. The structures of the plot transition networks varied according to the factors present in the detective stories. Therefore, an elaborate, human-like, and understandable creation of plot transitions can be enabled based on the detailed differences between plot transition networks.